ScaleDown 发布 3 个面向单一任务的 Constrained Decoding Models（Classify、Extract、Summarize），不做逐 token 生成，只对输入做一次 prefill 并从原文中直接选出答案。
Pulling a due date out of a 300-page PDF shouldn’t need a model that writes.
Decision models now get a 1M-token window, and they take images too.
ScaleDown released 3 small models (Classify, Extract, Summarize) for the simple, repetitive jobs that agents usually hand to an expensive LLM.
200x faster and 467x cheaper than frontier LLM APIs. $0.05 per million input tokens, and output is free.
The bigger change is that the same design now covers classify, extract and summarize.
The difference with regular LLMs is that these models don’t write their answer token by token. They read up to 1M tokens of text and images once and pick the answer straight from the input, so output is free, results stay inside your labels or source text
ScaleDown’s Constrained Decoding Models make one prefill pass over your input and select the answer from it: a label, a span with its source offset, or the passages that belong in a summary.
https://x.com/neal_k_patel/status/2109978626124898587/video/1
You already gave the LLM the answer. So why is it writing it out from scratch? Introducing a new AI model architecture: Constrained Decoding Models (CDMs) LLMs were built to write. Most production workloads just need an answer: fast, consistent, and affordable. CDMs are purpose-built for 1 task each. They read your input once and skip the token-by-token generative loop that makes LLMs slow and expensive. Not just a decision model, but a family of 3 CDMs live today: • Classify: define your labels and get back a calibrated probability + reason for each. Used for routing, tagging, and intent detection. • Extract: describe the entities in natural language and get back well-formed JSON with confidence scores + citations. Used for web scraping, document processing, and PII redaction. • Summarize: pass in long text, images, or PDFs and get back a consistent summary under 1,000 tokens. Used for context compaction, metadata generation, and transcript digests. Built for production: • 1M tokens of multimodal (text + image) input, 31x larger than Jev • 77.2% on DecisionBench vs Jev's 72% • 200x faster, 467x cheaper than frontier LLM APIs • $0.05/M input tokens. Output is free (yes, even for summarize) Live on your favorite platforms today: @OpenRouter @vercel_dev @LiteLLM @LangChain @TrustedRouter @itsArthurAI @useRouterPlus Get 50M input tokens for free (for the first 50 hours) at link below 👀在 X 查看被引用的帖子
来源：Rohan Paul · x.com